Background/Purpose: Concepts
of erosion pathogenesis in rheumatoid
arthritis (RA) have been based on radiographs, although MR images are much
better able to visualise erosions. Statistical shape models (SSM) allows bones
and erosions to be aligned, correcting for size and patient shape, allowing the
3D study of systematic effects. This study employed SSMs to visualise the
spatial distribution of erosions, and their probability of arising at a
particular location, in a population of early RA patients.

Methods: 90
baseline MR images were selected at random from an exploratory Phase 2
multicentre study (NCT01164579). Inclusion criteria were as follows: MTX-naïve
adult patients with early RA (duration ≤ 2 years); diagnosis using ACR
criteria, unequivocal evidence of radiographic erosion, and clinical evidence
of synovitis. 7 images were not analysed due to image quality.

Bones and erosions were manually segmented and independently
reviewed.  3D bone and erosion surfaces
were generated, and rigidly warped to the mean bone shape.  A population image was created; containing
the number of times that each voxel was found within an erosion. Voxels which
were present in >2 erosions were displayed along with the mean bones.
(Figure 1)

Results: Erosions
were present in low numbers of patients. In 18 patients there were no erosions
visible in MR despite reported radiographic erosions.  Over half had only 0 to 2 bones with an
erosion. Erosions exhibited a clear spatial pattern.  In the metacarpals, the erosions were most
prevalent in MCP2 and 3 (Table 2). In the wrist, there were typically 20
erosions per bone for half of the bones; the trapezium, trapezoid and proximal
metacarpals had ~10. Erosions occurred at consistent sites, with only one or 2
sites within each bone where the erosion was found in >2 patients.

Conclusion: This
is the first study to provide an accurate 3D visualisation of erosion sites in
early RA.  Erosions occurred at only a
small number of entheseal sites. Erosions in the metacarpals occurred in the
area containing the collateral ligament and capsular attachments. In the wrist,
the erosions were located primarily on the palmar side of the bone, and had a
more complex arrangement. Although multiple ligament, capsular and tendon
attachments exist in close proximity on each bone, we observed that erosion
sites were usually those in which there was a deep attachment site in
non-eroded patients. This suggests that the attachments which will generate an
erosion are those which experience the highest mechanical load.  Further careful study will be required to
pursue the detail of these anatomical locations, and will significantly help
our understanding of the pathogenesis of RA erosions.